Apparel quality control and AQL inspection

In short

Apparel quality control is the system of checks that runs from incoming fabric and trims, through cutting, sewing and finishing, to the packed carton, so that a shipment matches the buyer's approved sample, specification sheet and tolerances. Because 100% inspection of lakhs of garments is slow, costly and still not fully reliable, the industry accepts or rejects lots by attribute sampling plans indexed by Acceptance Quality Limit (AQL) — internationally ISO 2859-1, adopted in India as the IS 2500 series, and equivalently ANSI/ASQ Z1.4 in US buyer protocols. The topic is examined in Apparel Quality Assurance, Garment Manufacturing Technology and Quality Management papers (NIFT, NIFT-TEA, DKTE, SASMIRA, university B.FTech/B.Tech Textiles), and is asked in almost every QA interview at Shahi Exports, Gokaldas, Arvind, Raymond, Page Industries, Pearl Global, Orient Craft, Tirupur knitwear houses, and at third-party inspection firms (SGS, Intertek, Bureau Veritas, TÜV, QIMA).

Explained

Quality terms and the structure of apparel QC

• Quality in apparel = conformance to the buyer's specification: approved sample, tech pack, measurement chart with tolerances, shade band, and workmanship standard.
• Quality control (QC) detects defects in the product; quality assurance (QA) builds the system and process that prevents them. Inspection is appraisal; testing is measurement of a property.
• Inspection by attributes = each unit is judged conforming/non-conforming (used for garments). Inspection by variables = a measured value is recorded (used for measurements, GSM, strength; ISO 3951 series).
• Defect = any non-conformity to the specification; defective = a garment carrying one or more defects. One garment may hold several defects — hence DHU differs from defective %.
• Standard QC stages in an export unit:
  1. Pre-production: fabric and trim inspection, shade sorting, shrinkage/bow-skew test, fit/size-set, PP (pre-production) sample approval, pilot run.
  2. Cutting: marker and lay audit, spreading tension, ply count, cut-part audit, bundle ticketing, fusing bond check.
  3. Sewing (inline): operator self-check, roving/inline inspection every 1-2 hours per operation, traffic light card (green/yellow/red) at the machine, and end-of-line checking of 100% of sewn garments in most Indian export units.
  4. Finishing: pressing, thread trimming, measurement check, button pull, metal detection, final 100% or audit check.
  5. Internal final audit on packed cartons by AQL, before the buyer's or third party's inspection.

Defect classification and defect lists

• Critical defect: one that can cause injury or legal/safety failure — broken needle or metal in garment, loose drawcord on children's hoods, choking hazard trims, prohibited chemical finish, wrong care/fibre content label (consumer-law issue under the Legal Metrology packaged-commodity rules and BIS textile labelling practice). Usually zero tolerance (Ac = 0); a single critical defect fails the lot.
• Major defect: would likely cause the garment to be returned or marked down — open seam, broken stitch in a stress seam, hole, stain, measurement outside tolerance, shade mismatch between panels, misplaced pocket, faulty zip.
• Minor defect: a departure from the standard that an average consumer would accept — a small untrimmed thread, slight puckering, faint pressing mark, minor label skew.
• Defects are also grouped by source: fabric defects (slub, hole, barré, stripe bowing), sewing defects (skipped stitch, seam pucker, uneven topstitch, raw edge), pressing/finishing defects (shine, scorch, crease), packing defects (wrong size ratio, missing hang tag, wrong carton mark).
• Buyers issue their own graded defect list; an exporter's internal standard must be at least as tight.

Incoming fabric inspection — the 4-point system

• Fabric is checked on an inspection machine at 10-20 m/min under specified light (D65 or TL84 per buyer), usually 10% of rolls, biased towards new suppliers or doubtful lots.
• 4-point (ASTM D5430 four-point) penalty scale, applied to both warp and weft direction defects:

• Maximum 4 points per linear yard (or metre) however many defects occur; full-width or continuous defects are usually 4 points per yard of their run.
• A hole of any size is normally 4 points. Defects outside the usable (cuttable) width are ignored.
• Other systems: 10-point system (points scaled by length, different for warp and weft), Graniteville 78, Dallas system (defects per 10 yd). The 4-point system dominates because it is simple and buyer-recognised.
• Typical acceptance: ≤ 40 points per 100 yd² for wovens (many buyers use 20-28 for shirting and fine fabrics); limits are contractual, not statutory.

AQL sampling — the concept

• AQL (Acceptance Quality Limit): the worst process average, in percent non-conforming (or non-conformities per 100 units), that is considered acceptable as a long-run average. It is not a permitted defect level for a single lot and not a target.
• The plan is designed so that lots at the AQL have a high probability of acceptance (about 0.88-0.99 depending on code letter), while lots much worse than the AQL are very likely rejected.
• Apparel practice: AQL 2.5 for majors is the common export default; 1.5 for premium/branded and for shirts of high value; 4.0 for low-cost basics and for minors; critical defects at 0.
• Key risks: producer's risk (α) — a good lot (at AQL) rejected, typically 5%; consumer's risk (β) — a bad lot accepted, usually quoted at the LTPD/RQL point with β = 10%.
• Sampling plan types: single (one sample, Ac/Re) — almost always used in apparel; double (second sample if the first is inconclusive); multiple/sequential (several small samples). Double and multiple give smaller average sample sizes but are harder to administer on a packing floor.

Operating the plan: ISO 2859-1 / IS 2500 step by step

1. Define the lot — garments of one style/PO offered for inspection together, usually when ≥ 80% is packed.
2. Choose the inspection level — Level II is normal (default). Level I uses about 0.4× the sample (lenient, for trusted suppliers); Level III about 1.6× (tightened). S-1 to S-4 are special small-sample levels for destructive or costly checks.
3. Find the sample size code letter from lot size and inspection level, then the sample size n and, against the chosen AQL, the acceptance number Ac and rejection number Re (= Ac + 1 under normal single sampling).
4. Draw the sample at random across the whole lot — select about √(number of cartons) cartons (many protocols use √N + 1), spread across sizes, colours and production dates.
5. Inspect — workmanship on all sampled pieces; measurements on a sub-sample; plus carton/shipping marks, size ratio, barcode scan, trims, shade against the band, needle/metal detection, button pull (a common requirement for children's wear), and any buyer-specified field tests.
6. Decide — accept if observed defectives (or defects) d ≤ Ac; reject if d ≥ Re. A reject means 100% re-inspection and rework by the exporter, then re-offer (usually inspected at tightened level).
7. Switching rules (mandatory in ISO 2859-1, not optional):
   • Normal → Tightened: 2 of any 5 (or fewer) consecutive lots not accepted.
   • Tightened → Normal: 5 consecutive lots accepted on tightened inspection.
   • Discontinue inspection (stop accepting from that supplier until it improves) if 5 lots remain non-accepted on tightened inspection.
   • Normal → Reduced: 10 consecutive lots accepted, total non-conformities below the standard's limit number, production steady, and the responsible authority agrees. Reduced inspection uses about 40% of the normal sample size, with Re greater than Ac + 1.
   • Reduced → Normal: any lot not accepted, or production becomes irregular.
8. Record — inspection report with lot size, level, AQL, n, Ac/Re, defect tally by type and grade, photographs, measurement sheet, and the verdict (Pass / Fail / Hold-pending).

Inspection types in the shipment cycle

• Pre-production inspection (PPI): raw material and trim readiness before bulk cutting.
• Initial/first-article and TOP (top of production): the first pieces off the line, to catch a systematic error before it multiplies.
• DUPRO (during-production inspection): when 20-50% is stitched; the most useful commercially because rework is still cheap.
• Final random inspection (FRI) / pre-shipment inspection (PSI): AQL audit at ≥ 80% packed; this is the gate for shipment clearance.
• Container loading supervision (CLS): carton count, condition, stuffing, seal number.
• Factory/technical audit: systems audit of QC capability (distinct from a social-compliance audit).

Limits, tools and the Indian framework

• Limitations of AQL sampling: it judges a lot, not a process; a passed lot still contains defectives; small lots get disproportionately small samples; it cannot detect a defect the checker does not look for; attribute data hides how far out of tolerance a measurement is; it gives no root cause.
• Therefore AQL audit sits on top of process control: 7 QC tools (check sheet, Pareto of defect types, cause-and-effect/fishbone, histogram, stratification, scatter, control chart), p-chart/np-chart for defective proportion, DHU trend by operation, 5 Why, PDCA, Six Sigma DMAIC and poka-yoke (needle guard, stacker, template).
• Standards and bodies: ISO 2859-1 (attribute sampling by AQL) adopted as the IS 2500 series by BIS — quote the part/edition from the current BIS catalogue, as parts have been revised; ISO 3951 for variables sampling; ASTM D5430 for visually inspecting fabrics; ISO 9001:2026 is the current edition of the QMS standard; buyer protocols (Walmart, M&S, H&M, Primark, Decathlon, Target) override generic practice where stricter.
• Indian institutions: the Textiles Committee (Ministry of Textiles) operates inspection, testing and quality-certification services for textiles and garments; AEPC supports exporters on buyer compliance; NITRA, ATIRA, BTRA/TRA, SITRA, SASMIRA are the testing/research associations; BIS Quality Control Orders make certain textile/fibre products subject to mandatory certification — check the current QCO list, as the coverage and dates have been amended several times.
• Sharp-tool and needle policy: broken-needle register with all fragments taped against the issue slip, controlled scissors/cutters, and 100% metal detection of finished garments — a critical-defect control demanded by virtually every brand.
• Social/labour compliance (WRAP, SA8000, BSCI) and the four Labour Codes in force from 21 November 2025 govern the workplace, not product conformity — they are audited separately from AQL.

Key points

• AQL - the worst acceptable long-run process average, in % non-conforming; a quality *limit*, not a target or an allowance.
• ISO 2859-1 / IS 2500 series - the attribute sampling standard behind every garment AQL inspection; ANSI/ASQ Z1.4 is the equivalent used in US buyer manuals.
• Inspection levels - II is normal, I about 0.4× the sample (lenient), III about 1.6× (tightened), S-1 to S-4 for destructive checks.
• Code letter - read from lot size × inspection level; it fixes the sample size n (A=2 up to R=2000).
• Ac / Re - accept if defectives ≤ Ac, reject if ≥ Re; under normal single sampling Re = Ac + 1.
• Common apparel AQLs - critical 0, major 2.5, minor 4.0; 1.5 for premium or children's wear.
• Switching rules - 2 rejects in 5 lots → tightened; 5 clean lots → back to normal; 5 rejects on tightened → discontinue.
• Random sampling - draw from about √(cartons) spread over sizes, colours and production dates; a convenience sample invalidates the plan.
• Critical defect - safety or legal non-conformity (metal, drawcord, wrong fibre/care label); one occurrence fails the lot.
• DHU - defects per hundred units at end-line; the process metric, separate from the AQL verdict.
• 4-point system - 1/2/3/4 penalty points by defect length, maximum 4 points per linear yard, typical limit ≤ 40 points per 100 yd².
• DUPRO - during-production inspection at 20-50% stitched; cheapest point to correct a systematic fault.
• FRI at 80% packed - the standard pre-shipment gate; below that the lot is not representative.
• Measurement tolerance - the ± band in the tech pack; outside it is a major defect, and tolerances differ by point of measure.
• OC curve - probability of acceptance against true lot quality; it shows why a 5% defective lot can still pass.
• AOQ / AOQL - average outgoing quality and its worst value, the quality actually shipped when rejected lots are screened.
• Attributes vs variables - attribute plans count defectives (ISO 2859); variables plans use measured data and smaller samples (ISO 3951).
• Audit ≠ control - an AQL pass certifies a lot, not a capable process; Pareto, fishbone and p-charts fix the cause.

Formulas and examples

Sample-size and acceptance table, ISO 2859-1 single sampling, normal, General Inspection Level II (majors):

Example 1 - reading the plan and deciding. A Tirupur unit offers 5,000 T-shirts; buyer protocol is GIL II, AQL 2.5 major / 4.0 minor, critical 0.

• Lot 5,000 → code L, n = 200.
• AQL 2.5 → Ac 10, Re 11. AQL 4.0 → Ac 14, Re 15.
• Found: 12 majors, 9 minors, 0 critical.
• Majors 12 ≥ 11 → lot FAILED on majors (minors 9 ≤ 14 would have passed). Verdict: reject, 100% re-check and rework, re-offer at tightened level.

Example 2 - 4-point fabric inspection. A roll is 120 yd long, usable width 58 in. Defects found: 6 up to 3 in, 4 of 3-6 in, 3 of 6-9 in, 2 over 9 in.

• Points = (6×1) + (4×2) + (3×3) + (2×4) = 6 + 8 + 9 + 8 = 31 points.
• Points per 100 yd² = 31 × 3600 / (120 × 58) = 111,600 / 6,960 = 16.0.
• Against a 40-point limit the roll is accepted; against a tight 15-point shirting limit it would be rejected.

Example 3 - OC curve and AOQ. Same plan as Example 1 (N = 5,000, n = 200, Ac = 10). The true process is running at 5% defective.

• np = 200 × 0.05 = 10. Pa = P(X ≤ 10 | mean 10) ≈ 0.583 — a 5% defective lot still has a 58% chance of passing, double the AQL.
• AOQ = 0.05 × 0.583 × (5,000 − 200)/5,000 = 0.05 × 0.583 × 0.96 = 0.0280 → 2.80% defective shipped on average.
• ATI = 200 + (1 − 0.583)(4,800) = 200 + 2,002 = 2,202 garments inspected on average per lot.
• Lesson: AQL sampling protects against grossly bad lots, not against a mediocre process.

Example 4 - DHU, defective % and RFT. End-line checking of 400 shirts records 72 defects spread over 68 shirts.

• DHU = 72/400 × 100 = 18.
• Defective % = 68/400 × 100 = 17%; RFT = (400 − 68)/400 × 100 = 83%.
• Pareto of the 72: skipped stitch 26, seam pucker 18, uneven topstitch 12, stain 9, others 7 → the top two account for 44/72 = 61%, so corrective action goes to needle/thread tension and feeding at those two operations first.

Example 5 - sampling spread and inspection level. A lot of 8,400 trousers is packed in 350 cartons, 24 pcs per carton.

• Cartons to open ≈ √350 = 18.7 → 19 cartons (19 × 24 = 456 pcs available, enough for n = 200).
• At GIL II: n = 200, Ac 10 at AQL 2.5. At GIL III (supplier on tightened level after two failures): code M, n = 315, Ac 14.
• Note n rose 58% but Ac only 40% — level III is genuinely stricter, not merely bigger.

Where it is used

• Garment export houses (Shahi Exports, Gokaldas Exports, Orient Craft, Pearl Global, Texport, KPR Mill, Tirupur and Noida clusters) — inline, end-line and final AQL audits against buyer protocols.
• Domestic brands and retail supply chains — Raymond, Arvind, Page Industries, Aditya Birla Fashion, Trent/Zudio and Reliance Retail vendor QA, where AQL audits gate vendor payment.
• Third-party inspection and certification — SGS, Intertek, Bureau Veritas, TÜV SÜD, QIMA inspectors; the Textiles Committee's inspection and testing services.
• Fabric and mill QC — 4-point inspection and shade/shrinkage sorting at mills and processing houses before fabric is released to cutting.
• Job roles — QC Checker, Inline/Roving Inspector, QA Executive, Final Audit/Internal Auditor, Technical Auditor, Vendor Quality Manager, Merchandiser (who negotiates the AQL and defect list in the PO).
• Beyond apparel — the same ISO 2859-1 plans are used in footwear, home textiles, toys, electronics assembly and pharma packaging inspection.

Interview questions

• Q: What does AQL 2.5 actually mean? A: It is the worst long-run process average, 2.5% non-conforming, still treated as acceptable; lots at that level have a high chance of acceptance. It is not permission to ship 2.5% defectives in any one lot.
• Q: For a lot of 5,000 garments at GIL II and AQL 2.5, what are n, Ac and Re? A: Code letter L, sample 200, accept on 10, reject on 11.
• Q: Which standard governs AQL inspection and what is its Indian adoption? A: ISO 2859-1 for attribute sampling indexed by AQL, adopted by BIS as the IS 2500 series; ANSI/ASQ Z1.4 is the equivalent used in many US buyer manuals.
• Q: Difference between a critical, major and minor defect? A: Critical can injure the user or breach law (metal, unsafe drawcord, wrong care label) and is zero-tolerance; major would cause return or markdown; minor is a cosmetic departure an average consumer accepts.
• Q: Difference between DHU and defective percentage? A: DHU counts defects per 100 units, so one garment with three defects contributes three; defective % counts the garment once. DHU is always ≥ defective % for the same data.
• Q: When is a final inspection carried out, and why then? A: When at least 80% of the lot is packed, so the sample represents the whole production and carton/packing checks are meaningful.
• Q: Explain the switching rules. A: Two non-accepted lots out of five consecutive moves inspection from normal to tightened; five consecutive accepted lots on tightened returns it to normal; five lots still non-accepted on tightened means inspection is discontinued until the supplier improves.
• Q: A lot passes AQL but the buyer's store finds defects. Is the inspection wrong? A: Not necessarily — sampling always carries consumer's risk; a lot at twice the AQL can still pass about half the time. The fix is process control and tighter level/AQL, not blaming the audit.
• Q: Explain the 4-point system and its acceptance limit. A: Defects score 1-4 points by length, capped at 4 points per linear yard; points are normalised to 100 yd², and a typical contractual limit is 40 points per 100 yd² (tighter for fine fabrics).
• Q: Why do we use inspection level I or III instead of changing the AQL? A: The level changes sample size and hence the discriminating power of the plan at the same agreed AQL; the AQL itself is a commercial term in the PO and is not changed unilaterally.
• Q: How do you select the sample randomly on the packing floor? A: Open roughly the square root of the number of cartons, chosen across sizes, colours and production dates, and pull pieces from different layers — never only the top cartons kept ready by the factory.
• Q: What is AOQL and why does it matter? A: The maximum of the average outgoing quality curve — the worst average defect level that reaches the customer when rejected lots are 100% screened; it tells you the real quality the plan delivers.
• Q: What controls stop a broken needle reaching the customer? A: A needle-exchange policy with all fragments recovered and logged against the issue slip, controlled sharp tools, and 100% metal detection of finished garments, with the defect graded critical.

Common mistakes

• Treating AQL as a target or an "allowed defect quota" — quality teams then aim at 2.5% instead of zero, and the OC curve guarantees escapes.
• Reading the code letter from the sample size table instead of from the lot size and inspection level, or using Level II numbers after the buyer has moved the supplier to Level III.
• Assuming Re = Ac + 1 always; that holds for normal single sampling only, not for reduced inspection.
• Counting defects instead of defectives (or mixing the two) when the plan is indexed in percent non-conforming — this skews the verdict.
• Picking an accessible or factory-prepared sample, or only one size/colour, and calling it random.
• Inspecting at 40-50% packed and reporting it as a final inspection; and offering a lot without the measurement check, shade band or metal detection records.
• In the 4-point system, forgetting the 4-points-per-yard cap, counting defects outside the usable width, or using full width instead of cuttable width in the formula.
• Passing a lot that shows a critical defect because the major count is within Ac — one critical fails the lot.
• Quoting a withdrawn IS number or an old ANSI/MIL-STD designation; MIL-STD-105E is cancelled and superseded by ANSI/ASQ Z1.4 and ISO 2859-1.

Related topics

Textile testing and quality control · Social compliance in apparel: WRAP, SA8000 and amfori BSCI · Statistical quality control and the seven QC tools · Garment manufacturing technology: cutting, sewing and finishing · Apparel production planning and industrial engineering (DHU, RFT, line balancing) · Sustainability in textiles